This was my C# implementation of Parzen Window Density Estimation.
The following is my implementation of the same in Matlab:
parzen.m
function [retval] = parzen (matrix, dataPoint, variance)
[r c] = size(matrix);
A = ones(r, c).*dataPoint;
sub = matrix - A;
up = sub.^2;
dw = 2 * variance;
firstPart = 1/(sqrt(2*pi*variance));
retval = firstPart * exp((-1)*(up/dw));
What do you think about this implementation? Is it correct?
Indentation is automatically fixed in MATLAB by marking everything, and do Ctrl+I.
The common practice is to have the first level alligned all the way to the left. Click here to see a screenshot from the builtin MATLAB function ind2sub. I have made no edits to it.
matrix is not a very good variable name. I suggest you choose something else. I don't have a suggestion off the top of my head, but there are probably more descriptive names.
[r c] = size(matrix);
There's not much more to say, best practice is to separate output values with commas.
The following subtracts the value of datapoint from A. I'm assuming datapoint is a scalar, not a matrix.
[r c] = size(matrix);
A = ones(r, c).*dataPoint;
sub = matrix - A;
This can be simplified a lot. If you want to subtract one value from all elements of a matrix in MATLAB, you can just do:
sub = matrix - dataPoint;
If dataPoint is a matrix, of the same dimensions as matrix, then you can do:
sub = matrix - dataPoint
Yes, that was exactly the same as above. As long as the dimensions match, or one of the variables is a scalar, you can simply subtract one from the other.
I guess dw means down? If so, call it down, it's easier to understand, and is not very hard to write.
You mix camelCase and lowercase. I suggest you stick to one, and do dataPoint and returnValue.
You forgot end in the end. It works without it, but it is supposed to be there.
Is it correct?
I don't know. I don't know what it's supposed to do. I guess the simplest way to check it is to compare it to your C# implementation.
